Patch adds support for the following instructions:
SVE2 floating-point pairwise operations:
- FADDP, FMAXNMP, FMINNMP, FMAXP, FMINP
The specification can be found here:
https://developer.arm.com/docs/ddi0602/latest
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
| test/MC/AArch64/SVE2/faddp-diagnostics.s | ||
|---|---|---|
| 36 ↗ | (On Diff #201217) | Also p0/z is invalid predicate, e.g.: faddp z0.h, p0/z, z0.h, z1.h
// CHECK: [[@LINE-1]]:{{[0-9]+}}: error: invalid operandetc. |
| test/MC/AArch64/SVE2/faddp-diagnostics.s | ||
|---|---|---|
| 36 ↗ | (On Diff #201217) | I can add a test for that before merging. Managed to miss this for quite a few of the predicate instructions it seems. I have a patch downstream that improves the SVE predicate tests also as similar cases are missing there, I'll get that up after the SVE2 stuff is in. |